Alzheimer's disease (AD) is a neurodegenerative disease characterized by the aggregation of amyloid β-peptide (Aβ) into β-sheet-rich fibrils. Although plaques containing Aβ fibrils have been viewed as the conventional hallmark of AD, recent research implicates small oligomeric species formed during the aggregation of Aβ in the neuronal toxicity and cognitive deficits associated with AD. We have demonstrated that oligomers, but not monomers, of Aβ40 and Aβ42 were found to induce calcium signalling in astrocytes but not in neurons. This cell specificity was dependent on the higher cholesterol level in the membrane of astrocytes compared with neurons. The Aβ-induced calcium signal stimulated NADPH oxidase and induced increased reactive oxygen species (ROS) production. These events are detectable at physiologically relevant concentrations of Aβ. Excessive ROS production and Ca2+ overload induced mitochondrial depolarization through activation of the DNA repairing enzyme poly(ADP-ribose) polymerase-1 (PARP-1) and opening mitochondrial permeability transition pore (mPTP). Aβ significantly reduced the level of GSH in both astrocytes and neurons, an effect which is dependent on external calcium. Thus Aβ induces a [Ca2+]c signal in astrocytes which could regulate the GSH level in co-cultures that in the area of excessive ROS production could be a trigger for neurotoxicity. The pineal hormone melatonin, the glycoprotein clusterin and regulation of the membrane cholesterol can modify Aβ-induced calcium signals, ROS production and mitochondrial depolarization, which eventually lead to neuroprotection.
Alzheimer's disease (AD) is a devastating neurodegenerative disorder that causes a progressive and irreversible loss of higher brain functions and ultimately leads to severe cognitive decline in the elderly population. Genetically, early-onset familial forms of AD are caused by a mutation in one of the three genes which regulates amyloid β-peptide (Aβ) production: presenilin 1 (PS1), presenilin 2 (PS2) and amyloid precursor protein (APP) . However, most cases of AD are late-onset and most probably sporadic, a result of a combination of epigenetic and non-AD-related genetic factors. In addition to epigenetic and non-AD genetic factors, multiple low-impact genetic risk factors associated with sporadic forms of AD exist .
Histo-anatomically, AD is characterized by several neuropathological changes, namely amyloid deposition in senile neurofibrillary plagues and intracellular peptide tau deposition, which lead to neuronal loss .
Aggregated Aβ is the main histopathological finding in post-mortem brains of AD patients, but involvement of this peptide in the development of the disease is still subject to debate . The importance of studying the toxicity of Aβ is based on the finding that the aggregated peptide is highly toxic for neurons and is able to induce dementia, independently of the origin of the factors (genetic or sporadic) that have led to its aggregation.
It has been largely proposed that only neurons are lost through the course of the disease. Generally, microglia and astrocytes are the main immune cells that exert critical neuroprotective roles in the brain. They may effectively clear the toxic accumulation of Aβ at the initial stage of AD; however, their functions may be attenuated because of glial overactivation. Given the role of the astrocyte–neuron interaction for the normal functioning of the healthy brain, e.g. anatomical support, immunological response and signalling functions, it is very important to shed more light on the nature of astrocyte–neuron interplay in the context of perturbation of physiological conditions and neurodegeneration upon Aβ application.
We have found that a number of events have started from a Aβ-induced calcium signal in astrocytes that ultimately lead to neuronal death. In the present article, we review the interaction of neurons and astrocytes in pathological conditions under the influence of aggregated Aβ.
Aβ and calcium signal in astrocytes
The ability of Aβ to change intracellular calcium concentrations has been tested on a number of different models and cell types . Results vary depending on the cell type, type of amyloid and even on the company producing the peptide. Using primary co-cultures of hippocampal or cortical neurons and astrocytes, we have found that full peptides Aβ1–42 and Aβ1–40 and different short forms Aβ25–35, Aβ25–40 or Aβ1–28 in micromolar concentrations are able to induce [Ca2+]c changes in astrocytes, but not in neurons. This specificity for the astrocytic signal was not linked to any receptor or channel activation [6,7], but was induced by incorporation of Aβ into the membrane with pore formation, which was shown on artificial membranes and cells [8,9]. The ability of all Aβ to form a pore and stimulate a calcium signal in cells is dependent on the aggregation of the peptide [10–12]. Recently, we have found that oligomeric Aβ can stimulate astrocytes in very low concentrations and that only a single molecule of oligomeric Aβ1–42 can stimulate Ca2+-influx to astrocytes .
Selectivity of the astrocytes to Aβ-induced calcium signal in comparison with neurons is explained by their membrane composition with higher content of cholesterol that can be visualized by fluorescent indicator fillipin . Manipulation of the membrane cholesterol content in co-culture of neurons and astrocytes changes their ability to induce the calcium signal in response to Aβ. Thus incubation of the cells with cholesterol increases its content in the cellular membranes and enables Aβ to stimulate a [Ca2+]c rise in both neurons and astrocytes with significantly increased neurotoxicity. Reduction in the cholesterol in the membrane using statins inhibits incorporation of the Aβ into the membranes and consequently inhibits Aβ-induced calcium signalling in astrocytes and neurons, which eventually leads to neuroprotection . These findings link Aβ neurotoxicity to the other major familial risk of AD, i.e. loss of function in apoliprotein E (ApoE), a lipoprotein responsible for cholesterol transport [15,16]. However, the link between ApoE and membrane cholesterol content is not direct: cholesterol uptake in the brain is efficiently prevented by the blood–brain barrier, and mature neurons are thought to rely on glial cells for their cholesterol supply  and on the synthesis of cholesterol de novo. Therefore the Aβ-evoked calcium signal and neurotoxicity can be prevented by manipulating the membrane cholesterol content.
Conversely, Aβ-induced calcium signal can be blocked when omitting Ca2+ from the medium. Treatment of neurons and astrocytes with Aβ in medium without calcium not only prevented any toxicity for astrocytes, but also led to an even higher degree of protection for neurons [6,7]. This strongly suggests a direct link between Aβ-induced astrocytic calcium signal and neuronal cell death that is likely to be regulated by neuron–glia interactions. Such a high calcium signal in astrocytes can also affect neurons by the release of cytokines, interleukins and nitric oxide (NO) production [18–21].
Aβ induces ROS production in astrocytes
The role of reactive oxygen species (ROS) is increasingly recognized in aging and age-related diseases, including neurodegenerative disorders . Cells are protected against ROS damage by the antioxidant system and become vulnerable to free radicals only during induction of oxidative stress. Oxidative stress is caused by an imbalance in the redox state of the cell, either by overproduction of ROS or by dysfunction of the antioxidant systems. There is extensive literature suggesting that Aβ causes oxidative stress. The level of the major antioxidant in the brain, glutathione (GSH), is a key marker of the redox balance in the cells. Using live-cell imaging of GSH, which allows us to separate the GSH signal from neurons and astrocytes, we found a dramatic decrease in the GSH concentration in astrocytes and neurons induced by Aβ. This effect could be averted in both neurons and astrocytes by inhibition of the astrocytic calcium signal using Ca2+-free medium . It suggests that Aβ induced calcium-dependent ROS production in astrocytes, which consequently affects neurons. Different mechanisms were suggested about how Aβ induces ROS production, from stimulation of ROS production in mitochondria to the production of superoxide in Aβ aggregated in the presence of copper [22–24]. These processes are not dependent on Ca2+ and we found that NADPH oxidase, an enzyme expressed in astrocytes, can stimulate ROS production in response to stimulation of the [Ca2+]c elevation by calcium ionophores or stimulation of the P2Y receptors by ATP . Aβ can activate NADPH oxidase through different mechanisms; thus in neutrophils from AD patients, it can be stimulated through activation of P2X receptors , whereas in microglia, it can be stimulated through activation of the scavenger receptors CD36 . We have found that activation of NADPH oxidase in astrocytes, in response to Aβ, is independent of binding to the CD36 receptor, but more probably acts through the Ca2+-dependent activation of protein kinase Cβ [25,28]. Although expression of different forms of NADPH oxidase was also reported for neurons , Aβ did not activate ROS production in neurons, which can be explained by the absence of the Aβ-induced calcium signal [28,30]. Activation of NADPH oxidase in astrocytes is one of the key elements in the mechanism of Aβ-induced neurotoxicity; application of inhibitors of this enzyme is protective against neuronal and astrocytic cell death [28,30–33]. It is of particular importance that the Aβ-induced calcium signal is a trigger for NADPH oxidase activation. For example, inhibitors of NADPH oxidase, melatonin or antibodies for acetylcholine receptors failed to alter [Ca2+]c in astrocytes, but significantly reduced ROS production and protected both neurons and astrocytes against cell death [12,28,34]. Activation of NADPH oxidase induces oxidative stress that results in a decrease in GSH concentration in neurons and astrocytes. Neurons can be supported by astrocytes in GSH synthesis by provision of the precursors for production of GSH. Pre-incubation of cells with GSH precursors significantly reduced neuronal loss under Aβ exposure, confirming that neuron–astrocyte interaction is very important in the mechanism of Aβ neurotoxicity [28,30]. The decrease in the GSH level in neurons can be a result of the metabolic restriction and limitation of the GSH synthesis rather than a pure effect of oxidative stress. Another marker of oxidative stress is lipid peroxidation (LP), and it has been shown that there is an increase in levels of malondialdehyde and 4-hydroxynonenal in brain tissue and cerebrospinal fluid (CSF) of AD patients compared with controls . This has also been confirmed in brains from AD transgenic mouse . Such biochemical assays do not allow the separation of LP from neurons and astrocytes which can be achieved by fluorescent live imaging. Figure 1 shows the rate of LP in neurons and astrocytes under Aβ1–42 exposure. Aβ1–42 (2 μM), after a short delay, induced a significant increase in the rate of LP in astrocytes (Figure 1A). The effect of Aβ on neurons from co-culture was also significant, but it was less prominent than the effect observed in astrocytes (Figures 1A–1C). The increased rate of LP in both neurons and astrocytes under Aβ application was dependent on ROS production in astrocytic NADPH oxidase and could be blocked by the specific inhibitor diphenyleneiodonium (DPI; 0.5 μM). Thus the Aβ-induced calcium signal in astrocytes stimulates NADPH oxidase. ROS produced by this enzyme in astrocytes affects glial cells and neighbouring neurons and induced oxidative stress, which is abnormally elevated in neurons due to the limitation of metabolic support for GSH synthesis, i.e. precursors delivered by the glial cells.
Aβ induces lipid peroxidation in mixed cultures of astrocytes and neurons
Aβ induces loss of mitochondrial membrane potential
Application of Aβ to mixed cultures of neurons and astrocytes induced two types of mitochondrial depolarization in astrocytes in the first hours of incubation: a slow and progressive loss of ΔΨm, and a transient loss of mitochondrial membrane potential . Both types of the Aβ-induced ΔΨm changes require ROS produced by the NADPH oxidase and can be blocked by inhibitors of this enzyme. Fast and transient mitochondrial depolarization, under application of Aβ, is a result of induction of the mitochondrial permeability transition pore (mPTP) [28,30,37] (which requires calcium overload and ROS ). The ability of Aβ to stimulate opening of the mPTP was also demonstrated on isolated mitochondria [39,40]. The mPTP is a trigger for neuronal cell death after prolonged exposure of Aβ. Chemical (cyclosporin A ) or molecular (cyclophylin D) knockout  inhibition of the mPTP, or polyphosphate removal  reduces the number of dead neurons. The slow and progressive loss of mitochondrial membrane potential was prevented by NADPH oxidase inhibitors and also by antioxidants. Provision of mitochondrial substrates to Aβ-treated cells completely prevents this depolarization, suggesting that there is also ROS-induced damage of substrates supply to the mitochondria. We found that Aβ induces a slow and progressive NADH decrease in astrocytes that can be blocked by inhibitors of poly(ADP-ribose) polymerase-1 (PARP-1) .
PARP-1 is a DNA repair enzyme, activated by ssDNA breaks , which are induced by oxidative stress. A nuclear protein of 113 kDa, PARP-1 catalyses the formation of ADP-ribose from NAD and ribose, forming poly(ADP)-ribose (PAR) polymers [42,43]. The hyperactivation of PARP-1 has been implicated in neurodegeneration in a number of diseases with oxidative stress and impaired mitochondrial function [44,45]. Pre-incubation of neurons and astrocytes with inhibitors of PARP-1 protects neurons and astrocytes against Aβ-induced loss of ΔΨm and cell death . Activation of PARP-1 in neurons is very likely to be one of the main factors for induction of cell death after prolonged exposure of these cells to Aβ.
To summarize (as shown in Figure 2), oligomeric Aβ induces calcium signal in astrocytes, but not in neurons. Elevated [Ca2+]c stimulates NADPH oxidase to produce ROS. Increased ROS production induces oxidative stress in both astrocytes and neurons via activation of LP, and concomitant consumption of intracellular GSH and activation of PARP. This exacerbates the reduction in intracellular GSH in neurons due to insufficient delivery of GSH precursors from glia to neurons. Oxidative stress and limit of antioxidants induce neuronal cell death.
Schematic representation of glia–neuron interaction in AD
Astrocytes in Health and Neurodegenerative Disease: A joint Biochemical Society/British Neuroscience Association Focused Meeting held at Institute of Child Health, University College London, London, U.K., 28–29 April 2014. Organized and Edited by Jon Cooper (Institute of Psychiatry, King's College London, U.K.), Diane Hanger (King's College London, U.K.), Wendy Noble (King's College London, U.K.), Michael Sofroniew (University of California Los Angeles, U.S.A.), Alexei Verkhratsky (University of Manchester, U.K.) and Brenda Williams (Institute of Psychiatry, King's College London, U.K.).
This work was supported by the Leverhulme Trust.